The principle of cortical representations when thousands of real-life objects and categories are involved remains unclear. Here, we built a computational model of the human visual system by using a deep neural network and predicted the cortical responses to natural visual stimuli. In particular, we trained the model by using fMRI data obtained while subjects watched very long (>10 hours) natural movie stimuli that contained thousands of visual object categories. Based on the model, we systematically analyzed the activation patterns in the brain induced by different kinds of object categories. We found that the categorical information was represented by distributed and overlapping cortical networks, as opposed to discrete and distinct areas. Three cortical networks represented such broad categories as biological objects, non-biological objects, and background scenes. More fine-grained categorical representations in the brain suggest that visual objects share more (spatially) similar cortical representations if they share more similar semantic meanings.